Originally developed in the 1970s for the optimal control of systems governed by partial differential equations, the adjoint method has found several successful applications, e.g., in meteorology with large-scale 3D or 4D atmospheric data assimilation schemes, for carbon cycle data assimilation in biogeochemistry and climate research, or in oceanographic modelling with efficient adjoint codes of ocean general circulation models. Despite the variety of applications in these research fields, adjoint methods have only very recently drawn attention from the ocean acoustics community. In ocean acoustic tomography and geoacoustic inversion, where the inverse problem is to recover unknown acoustic properties of the water column and the seabed from acoustic transmission data, the solution approaches are typically based on travel time inversion or standard matched-field processing in combination with metaheuristics for global optimization. In order to complement the adjoint schemes already in use in meteorology and oceanography with an ocean acoustic component, this thesis is concerned with the development of the adjoint of a full-field acoustic propagation model for shallow water environments. In view of the increasing importance of global ocean observing systems such as the European Seas Observatory Network, the Arctic Ocean Observing System and Maritime Rapid Environmental Assessment (MREA) systems for defence and security applications, the adjoint of an ocean acoustic propagation model can become an integral part of a coupled oceanographic and acoustic data assimilation scheme in the future. Given the acoustic pressure field measured on a vertical hydrophone array and a modelled replica field that is calculated for a specific parametrization of the environment, the developed adjoint model backpropagates the mismatch (residual) between the measured and predicted field from the receiver array towards the source. The backpropagated error field is then converted into an estimate of the exact gradient of the objective function with respect to any of the relevant physical parameters of the environment including the sound speed structure in the water column and densities, compressional/shear sound speeds, and attenuations of the sediment layers and the sub-bottom halfspace. The resulting environmental gradients can be used in combination with gradient descent methods such as conjugate gradient, or Newton-type optimization methods tolocate the error surface minimum via a series of iterations. This is particularly attractive for monitoring slowly varying environments, where the gradient information can be used to track the environmental parameters continuously over time and space. In shallow water environments, where an accurate treatment of the acoustic interaction with the bottom is of outmost importance for a correct prediction of the sound field, and field data are often recorded on non-fully populated arrays, there is an inherent need for observation over a broad range of frequencies. For this purpose, the adjoint-based approach is generalized for a joint optimization across multiple frequencies and special attention is devoted to regularization methods that incorporate additional information about the desired solution in order to stabilize the optimization process. Starting with an analytical formulation of the multiple-frequency adjoint approach for parabolic-type approximations, the adjoint method is progressively tailored in the course of the thesis towards a realistic wide-angle parabolic equation propagation model and the treatment of fully nonlocal impedance boundary conditions. A semi-automatic adjoint generation via modular graph approach enables the direct inversion of both the geoacoustic parameters embedded in the discrete nonlocal boundary condition and the acoustic properties of the water column. Several case studies based on environmental data obtained in Mediterranean shallow waters are used in the thesis to assess the capabilities of adjoint-based acoustic inversion for different experimental configurations, particularly taking into account sparse array geometries and partial depth coverage of the water column. The numerical implementation of the approach is found to be robust, provided that the initial guesses are not too far from the desired solution, and accurate, and converges in a small number of iterations. During the multi-frequency optimization process, the evolution of the control parameters displays a parameter hierarchy which clearly relates to the relative sensitivity of the acoustic pressure field to the physical parameters. The actual validation of the adjoint-generated environmental gradients for acoustic monitoring of a shallow water environment is based on acoustic and oceanographic data from the Yellow Shark '94 and the MREA '07 sea trials, conducted in the Tyrrhenian Sea, south of the island of Elba. Starting from an initial guess of the environmental control parameters, either obtained through acoustic inversion with global search or supported by archival in-situ data, the adjoint method provides an efficient means to adjust local changes with a couple of iterations and monitor the environmental properties over a series of inversions. In this thesis the adjoint-based approach is used, e.g., to fine-tune up to eight bottom geoacoustic parameters of a shallow-water environment and to track the time-varying sound speed profile in the water column. In the same way the approach can be extended to track the spatial water column and bottom structure using a mobile network of sparse arrays. Work is currently being focused on the inclusion of the adjoint approach into hybrid optimization schemes or ensemble predictions, as an essential building block in a combined ocean acoustic data assimilation framework and the subsequent validation of the acoustic monitoring capabilities with long-term experimental data in shallow water environments.

Accurate knowledge of water content in seasonal snow can be helpful for water resource management. In this study, a distributed temperature index snow model based on temperature and precipitation as forcing data, is used to estimate snow storage in the Jollie catchment approximately 20km east of the main divide of the central Southern Alps, New Zealand. The main objective is to apply a frequently used assimilation method, the ensemble Kalman square root filter, to assimilate remotely sensed snow covered area into the model and evaluate the impacts of this approach on simulations of snow water equivalent. A 250m resolution remotely sensed data from Moderate Resolution Imaging Spectroradiometer (MODIS), specifically tuned to the study location was used. Temperature and precipitation were given on a 0.055 latitude/longitude grid. Precipitation was perturbed as input into the model, generating 100 ensemble members, which represented model error. Only observations of snow covered area that had less that 25% cloud cover classification were used in the assimilation precess. The error in the snow covered area observations was assumed to be 0.1 and grow linearly with cloud cover fraction up to 1 for a totally cloud covered pixel. As the model was not calibrated, two withholding experiments were conducted, in which observations withheld from the assimilation process were compared to the results. Two model states were updated in the assimilation, the total snow accumulation state variable and the total snow melt state variable. The results of this study indicate that the model underestimates snow storage at the end of winter and/or does not detect snow fall events during the ablation period. The assimilation method only affected simulated snow covered area and snow storage during the ablation period. That corresponded to higher correlation between modelled snow cover area and the updated state variables. Withholding experiments show good agreement between observations and simulated snow covered area. This study successfully applied the ensemble Kalman square root filter and showed its applicability for New Zealand conditions.

Mountain glaciers are expected to be the greatest contributor to sea level rise over the next century. Glaciers provide a good indicator of global climate and how to monitor their change is an increasingly important issue for climate science and for sea level rise forecasts. However, there has been little direct measurement of glacier volume change in New Zealand. This study explores the use of remotely sensed data for measuring glacier volume change from 1965 to 2006. Digital photogrammetric methods were used to extract topographic data of the Tasman Glacier from aerial photography and ASTER imagery for the years 1965, 1986, 2002 and 2006. SRTM C band data from 2000 were also analysed. Data were compared to an existing digital elvation model produced from the New Zealand Digital Topographic Database to test for their reliability. Using regression analysis, the data were filtered and points representing rock were used to correct points on the glacier ice for vertical bias. The quality of the data extracted from the aerial photography was good on rock and debris covered ice, but poor on snow. The data extracted from ASTER was much more reliable on snow in the upper glacier than the aerial photography, but was very poor in the lower debris covered region of the glacier. While the quality of the SRTM data is very high, there is a second order distortion present in the data that is evident over elevation differences. However, the overall mean difference of the SRTM rock from TOPODATA is close to zero. An overall trend could be seen in the data between dates. However, the 2006 ASTER data proved unreliable on the debris covered section of the glacier. Total volume change is therefore calculated for the period between 1965 and 2002. The data show a loss of 3:4km³ or 0:092km³ per year, an estimated 6% of the total ice in New Zealand. This is compared to estimates using the annual end of summer snowline survey between 1977 and 2005 of 1:78 km³, or 0:064km³ per year. The spatial resolution of ASTER makes high temporal resolution monitoring of volume change unlikely for the New Zealand glaciers. The infrequency of aerial photography, the high cost and vast time involved in extracting good quality elevation data from aerial photography makes it impractical for monitoring glacier volume change remotely. However, SRTM and other radar sensors may provide a better solution, as the data do not rely heavily on user processing.

Phymatotrichum root rot is a fungal disease with a host range that includes many economically important crops in the southwestern United States and Mexico. While it has been studied since the late nineteenth century, ecological relationships of the disease, particularly those related to its distribution and dispersal, are not understood. Combined ground radiance sampling and aerial photographic interpretation was employed to study the distribution of Phymatotrichum root rot in cotton. Radiometric ground sampling showed that diseased cotton has a characteristic spectral signature that is significantly different from healthy cotton at visible wavelengths. Micro-scale examination of distribution within fields utilized multitemporal photography, both within season (1983) and for four seasons (1979-1982), revealed that the disease spreads during a season, but is not recurrent in many cases between years. Meso-scale mapping employed multitemporal photography to map distribution during a four-year period. When compared to mapped soil units, these data revealed a significantly non-random relationship between the diseased areas of fields and fine-textured soil units that may be based on moisture-holding potential. A yield analysis was also preformed using Thematic Mapper Simulator data and computer analysis.

Egypt witnesses an increasing population growth concomitant with limited water and agricultural land resources. The objectives of this study were to utilize remotely sensed data for the inventory of agricultural lands in the Nile Delta, monitoring spatial and temporal variations in agricultural lands and quantifying agricultural land losses due to urbanization. Inventory of agricultural lands was designed using two approaches: thresholding and linear mixture analysis. We utilized 12 images from the Landsat satellite: 4 from Multi-Spectral Scanner (1972), 4 from Thematic Mapper (1984) and 4 from Thematic Mapper (2003) covering the entire Nile Delta. In addition, a set of 480 NDVI images were obtained from the Advanced Very High Resolution Radiometer (AVHRR) sensor that cover the period 1984-2003. Landsat images were subjected to atmospheric, radiometric and geometric corrections as well as image mosaicking. Normalized Difference Vegetation Index (NDVI) was applied and thresholding for agricultural land cover revealed that the areal extent of agricultural lands was 3.68, 4.32 and 4.95 million acres (one acre = 0.96 Egyptian Feddan) in 1972, 1984 and 2003, respectively. Linear mixture analysis of the AVHRR-NDVI with the TM-NDVI images showed that agricultural lands approached 4.11 and 5.24 million acres in 1984 and 2003, respectively. Using multitemporal Principal Component Analysis (PCA) for the TM and AVHRR images proved that reclamation activities were mostly along the western margins of the Nile Delta. Spatio-temporal analysis showed that middle delta has the highest agricultural vigor compared with the margins. Agricultural land loss was estimated in some cities within the delta as well as in Greater Cairo area. We studied the land cover classification and change in Greater Cairo area based on 5 Landsat images acquired in 1972, 1984, 1990, 1998 and 2003. Agricultural lands lost 28.43% (32,236 acres) between 1972 and 2003 with an annual loss of 1040 acres. Agricultural lands on the peripheries of Cairo and its satellite towns were the most vulnerable areas. Soil salinization was another limiting factor for land reclamation. The main conclusion confirms that remote sensing is an accurate, efficient and less expensive tool for the inventory and monitoring agricultural land change in Egypt.

A three-phase study was performed to examine abiotic and biotic metrics at southern pine beetle infestation sites in northern Georgia in 2002 to find early indicators that can be leveraged by forest managers to mitigate the effects of future outbreaks: creation of a 2003 Final Impact Map, determining if MODIS MOD13Q1 EVI 16-day image composites can distinguish differences in biomass indicators among healthy and infested loblolly pine and hardwood forests, and creation of an Infestation Risk Map derived from significant climate and physical variables at known infestation sites.Three land cover classification techniques (change vector analysis, enhanced wetness differencing index and standard land cover classification analysis of Landsat 5 TM) were compared to determine which would provide the best estimate of final infestation damage. Classification accuracy results indicated that the latter provided the most reliable site damage information and it became the reference map against which outbreak model results were compared.Using time series analysis of MODIS composites acquired March 2000 - December 2006 to measure 11 phenology metrics for infested and healthy loblolly and hardwood stands showed that the imagery differentiated between forest classes. Results indicated the lowest base vegetation biomass in 2001 for infested loblolly, relative to healthy loblolly, with many metrics trending towards hardwood values following infestation.Abiotic influences included those related to landscape position and climate. Statistical testing showed increased beetle success: 1) along ridge tops at maximum solar exposure, 2) in areas with canopy density>60%, 3) in areas experiencing cooler summers and warmer winters, and 4) where precipitation was significantly lower at infested sites in the 2 years preceding outbreak.The Infestation Risk Map was developed from significant physical and climate indicator variables using the fuzzy theory modeling approach. Comparison of model output to infestation sites resulted in Chi-squared and Cramér's V values of 55.4 and 0.16, respectively, indicating that infestation risk distributions strongly paralleled site infestation. Comparison of model output and low, medium and high infestation density clusters resulted in Chi-squared and Cramér's V values of 241.24 and 0.66, respectively, indicating a more substantive relationship between infestation density and risk classes.

New forms of disturbance, and alteration of current disturbance regimes in arid and semiarid ecosystems, have resulted in the modification and degradation of large regions. This research explores vegetation response as a consequence of two different disturbance events in the southwestern US and northwestern Mexico. This topic was explored in this dissertation utilizing remotely sensed geospatial information in three separate studies.The first study explores the development of methods to assess the effectiveness of pre-fire restoration efforts, by evaluating vegetation response as a function of local environmental variables. Here I evaluated three fire locations at Bandelier National Monument (New Mexico). My models explain post-fire vegetation response as a function of environmental inputs and pre-fire site conditions (restored, unrestored and control areas). However, further analysis will be needed to better understand the effect of pre-fire restoration techniques on post-fire vegetation response.My second study explores the development of monitoring practices using remotely sensed data to assess land cover dynamics through time. The study area was the arid agro-ecosystem of La Costa de Hermosillo (LCH) in northwestern Mexico. My results show a continuous tendency towards a decrease in agriculture from 1988 until 2009. Detailed change detection demonstrates high rates of change from agriculture to other land cover classes in areas with dense agricultural developments. Implementation of these monitoring protocols would help with the application of restoration practices.The third study we used remote sensing time series data to assess phenological trends and variability among land cover types in relation to climatic variability within communities present in a heavily impacted agro-ecosystem (LCH). My analysis comprised three different agricultural land cover types including abandoned agricultural fields, and four additional natural land cover classes. I found that productivity has not increased in abandoned fields (since abandonment). Furthermore, I found that the models developed in this study significantly explain phenological variability as a function of climatic variability.These studies suggest that the use of remote sensing tools could effectively contribute to our ability to monitor vegetation dynamics in arid ecosystems. The implementation of methodologies generated in this work would significantly inform managers in decision making processes.

The Cienega de Santa Clara is a 5600 ha, anthropogenic wetland in the delta of the Colorado River in Mexico. It is the inadvertent creation of the disposal of brackish agricultural waste water from the U.S. into the intertidal zone of the river delta in Mexico, but has become an internationally important wetland for resident and migratory water birds. The marsh is dominated by Typha domengensis with Phragmites australis as a sub-dominant species in shallower marsh areas. The most important factor controlling vegetation density was fire. The second significant (P<0.01) factor controlling NDVI was flow rate of agricultural drain water from the U.S. into the marsh. Reduced summer flows in 2001 due to canal repairs, and in 2010 during the YDP test run, produced the two lowest NDVI values of the time series from 2000 to 2011 (P<0.05). Salinity is a further determinant of vegetation dynamics as determined by greenhouse experiments, but was nearly constant over the period 2000 to 2011, so it was not a significant variable in regression analyses. Evapotranspiration (ET) and other water balance components were measured in Cienega de Santa Clara; we used a remote sensing algorithm to estimate ET from meteorological data and Enhanced Vegetation Index values from the Moderate Resolution Imaging Spectrometer (MODIS) sensors on the Terra satellite. We used Landsat NDVI imagery from 1978-2011 to determine the area and intensity of vegetation and to estimate evapotranspiration (ET) to construct a water balance. Remote sensing data was supplemented with hydrological data, site surveys and literature citations. The vegetated area increased from 1978 to 1995 and has been constant at about 4200 ha since then. The dominant vegetation type is Typha domingensis (southern cattail), and peak summer NDVI since 1995 has been stable at 0.379 (SD = 0.016), about half of NDVI(max). About 30% of the inflow water is consumed in ET, with the remainder exiting the Cienega as outflow water, mainly during winter months when T. domingensis is dormant.

A fundamental knowledge gap in understanding land-atmosphere interactions is accurate, high-resolution soil properties. Remote sensing and spatial modeling techniques can bridge the gap between site-specific soil properties and landscape variability, thereby improving predictions of soil attributes. Three studies were completed to advance soil prediction models in semiarid areas. The first study developed a soil pre-mapping technique using automated image segmentation that utilized soil-landscape relationships and surface reflectance to produce an effective map unit design in a 160,000 ha soil survey area. Overall classification accuracy of soil taxonomic units at the suborder was 58 % after including soil temperature regime. Physical soil properties were not significantly different for individual transects; however, properties were significantly different between soil pre-map units when soils from the entire study area were compared. Other studies used a raster approach to predict physical soil properties at a 5 m spatial resolution for a 6,265 ha area using digital soil mapping. The second study utilized remotely-sensed auxiliary data to develop a sampling design and compared three geostatistical techniques for predicting surface soil properties. Ordinary kriging had the smallest prediction error; however, regression kriging preserved landscape features present in the study area and demonstrated the potential of this technique for quantifying variability of soil components within soil map units. The third study applied quantitative data from soil prediction models in study 2 and additional models of subsurface properties to a pedotransfer function for predicting hydraulic soil parameters at the landscape scale. Saturated hydraulic conductivity and water retention parameters were used to predict water residence times for loss to gravity and evapotranspiration across the landscape. High water residence time for gravitational water corresponded to both low drainage density and high clay content, whereas high residence of plant available water was related to increased vegetation response. These studies illustrate the utility of digital soil mapping techniques for improving soil information at landscape scales, while reducing required resources. Resulting soil information is useful for quantifying landscape-scale processes that require constraint of spatial variability and prediction error of soil properties to better model hydrological and ecological responses to climate and land use change.

Dryland ecosystems provide a unique opportunity to study the effects of water limitation on ecosystem activity. The sensitivity of these systems to small inputs of moisture is well-documented, but the expression of water limitation in terms of carbon dioxide flux between the ecosystem and atmosphere remains unclear. Applying a simple conceptual approach to soil moisture dynamics, patterns in carbon flux become clear. Release of carbon dioxide via respiration is primarily driven by moisture in the shallow soil, and differences in respiration rates among plant functional types are only evident after controlling for soil moisture. Alternatively, carbon uptake by a semiarid shrubs ecosystem is largely driven by the availability of deep soil moisture. This link to deep soil moisture improves spatial scaling of gross and net carbon uptake using remote sensing data. Lateral redistribution of moisture on the landscape connects readily observed physical features, namely topography, to ecosystem function, but redistribution is generally not considered in carbon models. A simple runoff scheme coupled to a conceptual model for carbon flux demonstrates the high degree of spatial heterogeneity in carbon dioxide flux resulting from moisture redistribution. The importance of redistribution in carbon modeling is highlighted by interannual variability in modeled carbon fluxes under different rainfall characteristics (event size, event duration, interstorm duration). The links between hydrology and ecology across spatial scales become clearer when topographically-based moisture distribution is used as an organizing variable. In all, this research identifies new avenues for research where moisture dynamics are of central interest in dryland ecohydrology.